Technical challenges in quantifying the risk of measles outbreaks


Sebastian Funk (@sbfnk)
https://epiforecasts.io

14 September, 2023

Source: WHO

Can we quantify the risk of measles outbreaks (and if yes, using which data/models)?

\(R_0\) can span wide range

Guerra et al., Lancet Inf Dis, 2017

How do we best estimate susceptibility?

Population-level susceptibility depends on:

  1. Vaccination history
  2. Disease history
  3. Migration

(1) and (2) are in principle reported via coverage and case data

Lexis diagrams

Abrams et al., Am J Epid, 2014

Model: Vaccination → immunity

  1. map year of report to current age (e.g., 1st dose at 1y of age in 1990, 34y old now)
  2. assume 2nd doses assumed given according to schedule and distributed randomly
  3. assume 90% vaccine efficacy at each dose
  4. assume SIAs target unvaccinated

Model: Disease → immunity

  1. assume cases distributed by agelike cases in the EUR region in the last 10 years Example: 10,000 cases reported in 1995 4.6% (460) in 6y olds, 25y old now
  2. assume 50% of infections reported

Comparison to serology (ESEN2)

Why are the estimates so different?

  1. Serology wrong
  2. Vaccination data wrong
  3. Case data wrong
  4. Model wrong

Immunity profiles vs. outbreaks

Does serology predict outbreaks?

ESEN2 seroprevalence study

  1. Standardised measles seroprevalence in 17 European countries + Australia, conducted 1996-2004.
  2. Combination of residual and population random sampling (Andrews, 2008, Bull World Health Organ)

ESEN2 vs outbreaks

NHANES vs outbreaks

Quantifying risk in the absence of serology

Which variables explain measles growth rates in France?

Robert et al., BMC Med, 2022

Challenges in quantifying outbreak risk

  • General challenges
    • Local context and data
    • Ever more fine-grained approaches needed towards elimination
  • Serology for quantifying susceptibility
    • Representativeness
    • Scale
  • Models for quantifying susceptibility
    • Value, meaning and reliability of input data
    • Quantifying outbreaks in heterogeneous surveillance systems